The Contribution of Dutch GLOBE Schools to Validation of Aerosol Measurements from Space

The Contribution of Dutch GLOBE Schools to Validation of Aerosol Measurements from Space Joris de Vroom Vrije Universiteit Amsterdam October 23

Index List of Acronyms...3 1 Introduction...4 2 GLOBE Sun photometer measurements...9 2.1 Instrument description...9 2.2 Physical principles...1 2.3 Measurement method...11 2.4 Assumptions...12 2.4.1 Relative air mass and vertical distribution...12 2.4.2 Finite bandwith and BBL...14 2.5 Conclusion...17 3 Instrument Calibration...18 3.1 Langley analysis of RG2-47...18 3.2 Relative calibrations...19 3.3 Conclusion...19 4 Algorithm...2 4.1 AOT Calculation...2 4.2 Algorithm comparison...23 4.3 Error analysis...24 4.3.1 Random Errors...24 4.3.2 Systematic errors...28 4.4 Conclusion...28 5 Validation of GLOBE AOT measurements...29 5.1 Validation of GLOBE Sun photometer AOT measurements with SPUV AOT measurements....29 5.2 Validation of GLOBE AOT measurements by undergraduate students with AERONET AOT measurements....32 5.3 Conclusion...33 6 MODIS AOT Validation...35 6.1 The MODIS instrument...35 6.2 MODIS Algorithm...35 6.3 Validation results...38 6.4 Conclusion...43 7 Conclusion and Outlook...44 REFERENCES...46 Acknowledgements...47 2

List of Acronyms AERONET Aerosol Robotic Network AOT Aerosol Optical Thickness BBL Beer -Bougeur-Lambert FEL Fysisch en Electrisch Laboratorium (Physics and Electronics Laboratory) GLOBE Global Learning and Observations to Benefit the Environment GOME Global Ozone Monitoring Experiment IR Infra-Red KNMI Koninklijk Nederlands Meteorologisch Instituut (Royal Dutch Meteorological Institute) LED Light Emitting Diode LUT Look Up Table MODIS Moderate Resolution Imaging Spectroradiometer NASA National Aeronautics and Space Administration OMI Ozone Monitoring Instrument PI Principle Investigator SCIAMACHY Scanning Imaging Absorption Spectrometer for Atmospheric Chartography SME Stichting Milieu Educatie (Foundation for Environmental Education) SPUV Sun Photometer Ultra Violet TNO Nederlandse Organisatie voor Toegepast Natuurwetenschappelijk onderzoek (Netherlands Organization for Applied Scientific Research) TPD Technische Physische Dienst (Technical Physical Service) UT Universal Time VIS Visible 3

1 Introduction In the past century, the mean surface temperature of the Earth has increased by.6 according to the International Panel of Climate Change [IPCC, 21]. Figure 1.1 shows the mean Earth s surface temperature since 186. The red bars are the mean annual temperature deviation from the 1961 199 average and the red line is a 1 year average. The black line in Figure 1.1 shows a distinct trend upward. Since the industrial revolution, large-scale industrious human activities have accounted for an increase in greenhouse gas concentrations in the atmosphere. Greenhouse gases like carbon dioxide (CO 2 ) and methane (CH 4 ) absorb infrared radiation, thereby influencing the radiation balance and warming the Earth. Furthermore, human activities have accounted for a large increase in small solid or liquid atmospheric particles, called aerosols. The effect of aerosols on the radiation budget is complicated. Figure 1.2 [IPCC, 21] shows the level of understanding of the radiative effects, called radiative forcing, resulting from an increase in aerosol and greenhouse gas concentrations [IPCC, 21]. The radiative forcing caused by aerosols is still very poorly understood, but might even be as large (but negative) as the radiative forcing caused by greenhouse gasses. The first mechanism by which aerosols influence our climate system is by reflecting and absorbing Solar radiation and infrared thermal radiation from the Earth s surface. This is called the aerosol direct effect. Some aerosol species, like for example black carbon, absorb radiation at long wavelength, thus warming the Earth s climate system. Other species, like for example desert dust, effectively scatter Solar radiation, thus increasing the planetary albedo and cooling the Earth s climate system. Aerosol direct forcing is currently considered to give a cooling effect [Kaufman et al, 21], although some aerosol types, like black carbon, may give a warming effect. The second mechanism by which aerosols influence our climate system is by using as condensation nuclei for water vapor, thereby enhancing the process of cloud forming, resulting in more clouds. Furthermore, these clouds consist of more cloud droplets, which are smaller in size. This increases both the reflectivity and lifetime of clouds. This is called aerosol indirect effect. Clouds reflect Solar radiation, giving a cooling effect, and they reflect infrared radiation from the Earth s surface, giving a warming effect. The net effect of the aerosol indirect effect is currently considered to be cooling the Earth s surface [Kaufman et al., 22]. The effects of aerosols in our climate system described above illustrate the need to monitor aerosols. Besides the effect on climate, other reasons for the need to monitor aerosols are that high aerosol concentrations in urban regions can cause smog, which may lead to human health problems, and that aerosols affect the chemical composition of the atmosphere by the alteration of photolysis rates and by direct chemical interaction with gasses [Kaufman et al, 22]. Aerosols can be grouped into five categories: dustlike soil, soot, sulfate, sea salt and organic aerosols [Liou, 22]. Dustlike soil and sea salt aerosol particles have a typical diameter larger than 1 m while soot, sulfate and organic aerosol have a typical diameter smaller than 1 m. Aerosol concentrations are highly variable in space and time, caused by the relatively short lifetime of an aerosol particle, and the differences in aerosol sources. Examples of aerosols sources are forest fires, human industrial activities and sand storms. Furthermore, it is difficult to distinguish anthropogenic aerosols from natural aerosols. Except for marine aerosol, all types of aerosols can have both natural and anthropogenic sources. However, it is possible to estimate the anthropogenic contribution to the total aerosol load using satellite data, aerosol models and information on fire practices and agricultural and industrial activities. 4

Figure 1.1. Variations of the earth s surface temperature for the past 14 years, deviation from the 1961 199 average. Figure adopted from IPCC, 21. Figure 1.2. The global mean radiative forcing of the climate system for the year 2 relative to 175. Figure adopted from IPCC, 21. Since aerosols influence the climate system of the Earth and since there are still large uncertainties about their radiative effects it is necessary to monitor aerosols. This is done on a local basis by solar radiance measurements with Sun photometers, which measure the light extinguished by aerosols, expressed in aerosol optical thickness (AOT). These Sun photometers have the advantage that they are able to produce continuous AOT time series with high accuracy. However, such time series are valid only for the fixed measuring location. A network of Sun photometers is the Aerosol Robotic Network (AERONET), with over a hundred Sun photometers, stationed all over the world. AOT measurements between various AERONET instruments on different locations show large differences. Therefore, measurements on a global scale are essential. A way to measure AOT on a global scale is by using satellite instrument measurements. An advantage of satellite instrument measurements is that global coverage can be achieved within a few days. However, difficulties with satellite measurements may be expected due to instrument degradation in the harsh outer space environment and the fact that the instrument cannot be approached after launch. Therefore it is essential that the accuracy and precision of satellite instrument measurements are checked by comparisons to ground-based measurements throughout the entire satellite mission. This is called validation. In order to achieve a high accuracy of validation, it is desirable to validate as many satellite ground pixels as possible. Validation with expensive, professional Sun photometers is limited to a relatively small amount of ground pixels for satellite instruments with a spatial resolution of 13 x 24 km 2, such as the Ozone Monitoring Instrument (OMI) (Figures 1.3 and 1.4). The contribution of the GLOBE Aerosol Monitoring Project to satellite validation is investigated in this report. The GLOBE Aerosol Monitoring Project has the potential to supply a tight network of ground-based AOT measurements that can validate many satellite pixels over land, thereby improving satellite validation. The GLOBE Aerosol Monitoring Project The GLOBE (Global Learning and Observations to Benefit the Environment) program is an international science and education program. It started in 1995 by initiative of Al Gore, the then vicepresident of the United States, and is currently running in 12 countries. GLOBE is a partnership between the United States and the 12 other countries. The goal of the GLOBE program is involving primary and secondary school students in practical science by taking scientifically valid measurements. 5

Figure 1.3. The Ozone Monitoring Instrument, autumn 21, at the beginning of testing period at TNO-TPD. Picture by TNO-TPD. Figure 1.4. OMI on EOS-AURA (artist impression). The GLOBE Aerosol Monitoring Project is coordinated by David Brooks, from the Drexel University in Philadelphia. The project involves validation of satellite AOT measurements with AOT measurements by undergraduate students using a handheld Sun photometer, developed for the GLOBE program [Brooks and Mims, 21]. The Dutch division of GLOBE is represented by GLOBE Netherlands. Over a hundred schools are involved in cloud observations, phenology and measuring meteorological parameters, acidification and aerosols. The Dutch GLOBE Aerosol Monitoring Project is a co-operation between SME (Foundation for Environmental Education) and KNMI (Royal Dutch Meteorological Institution). SME accounts for the organization and contact with schools and KNMI is responsible for data processing, quality control and satellite validation. KNMI has experience in satellite validation with the satellite instrument GOME (Global Ozone Monitoring Experiment). Furthermore, KNMI is Principle Investigator (PI) for the validation of the Dutch-German-Belgium satellite instrument SCIAMACHY (Scanning Imaging Absorption Spectrometer for Atmospheric Chartography) and is PI for OMI. The goals of the Dutch GLOBE Aerosol Monitoring Project for KNMI are threefold: 1. Validation of satellite aerosol optical thickness measurements. 2. Public outreach for Dutch satellite instrument missions such as SCIAMACHY and OMI. 3. Involve students in practical science. To illustrate the potential of the GLOBE Aerosol Monitoring Project, AOT measurements at 44 nm by the satellite instrument MODIS (Moderate Resolution Imaging Spectroradiometer) from April 8 23 at 1: UT (Universal Time) are plotted in Figure 1.5. The white areas indicate regions without a reliable MODIS measurement, for example over sea or clouds. The two black dots in Figure 1.5 represent the two professional Sun photometers in the Netherlands at TNO-FEL (Netherlands Organization for Applied Scientific Research - Physics and Electronics Laboratory) in The Hague and at KNMI in De Bilt. The grey dots represent schools that currently participate in the GLOBE Aerosol Monitoring Project. Table 1.1 shows the names of the nine schools and their location. Figure 1.5 shows that the Dutch GLOBE school network improves the potential of validating satellite measurements over the Netherlands. Since nine schools are interested in joining the GLOBE training in September 23 at KNMI, the number of schools is expected to get larger. The Bernard Nieuwentijt college, Pascal college, Alkwin college, SG. Tabor and Goois lyceum are urban regions, in or close to Amsterdam. Christelijk college De Populier and Zwin college are located at a very interresting locations, since they are both coastal regions and relatively close to Rotterdam (51.7 N, 4.2 E) which is a very industrious area, and various types of aerosol may be expected at different wind directions. Furthermore De Populier is close to the AERONET instrument at TNO- FEL in The Hague. The Mozaiek college and Ichthus college are located at less industrious regions. The region around Amsterdam shows relatively good coverage. SME is looking for more schools in the East, South-East and North-East to generate a more homogeneous coverage over the Netherlands. 6

Figure 1.5. GLOBE potential for satellite validation. MODIS AOT at 44 nm over the Netherlands on 23-4-8 School name School location Latitude (North) Longitude (East) Participating Since Bernard Nieuwentijt College Amsterdam 52.38 4.93 1-22 Pascal College Zaandam 52.46 4.83 1-22 Christelijk Collgede de The Hague 52.5 4.16 1-22 Populier Mozaiek College Arnhem 51.98 5.92 1-22 Alkwin College Uithoorn 52.25 4.82 3-23 SG. Tabor, Locatie Oscar Hoorn 52.7 5.1 3-23 Romero Goois Lyceum Bussum 52.16 5.1 3-23 Zwin College Oostburg 51.32 3.49 3-23 Ichthus College Kampen 52.5 5.9 3-23 Table 1.1. Participating schools and locations. In reference to aerosols and its role in the climate system, the need to measure aerosol with satellite measurements and the potential of the Dutch GLOBE Aerosol Monitoring school network, the main questions that is to be answered in this paper is: Can the Dutch GLOBE school network be used to validate satellite instruments? In order to answer this question, the following sub questions are formulated and answered in this paper: 1. Can GLOBE Sun photometer measurements be used for validation in a theoretical way? This is answered by looking at the instrument and its measuring method, development of an algorithm and analysis of error sources. 7

2. Can GLOBE Sun photometer measurements be used for validation in a practical way? This is answered by comparing GLOBE Sun photometer measurements done at KNMI with measurements by a professional Sun photometer. 3. Can GLOBE Sun photometer measurements by undergraduates be used for validation? This is answered by comparing Sun photometer measurements done by undergraduates with measurements by a professional Sun photometer. Finally, the results are applied to validation of MODIS AOT measurements over the Netherlands by the Dutch GLOBE aerosol monitoring school network. 8

2 GLOBE Sun photometer measurements The GLOBE Aerosol Monitoring Project involves aerosol optical thickness (AOT) measurements with light emitting diode (LED)-based GLOBE Sun photometers by undergraduate students. The use of a LED as a detector is what distinguishes the GLOBE Sun photometer from professional Sun photometers. The relatively wide spectral response of the LED effects the interpretation of results from GLOBE Sun photometer measurements. The purpose of this chapter is to investigate the possibility of accurately measuring AOT with the GLOBE Sun photometer. The GLOBE Sun photometer instrument is described in section 2.1. The physical principle that underlies the measuring method is the Beer-Bougeur-Lambert (BBL) law, which is introduced in section 2.2. The measuring method of the GLOBE Sun photometer is discussed in section 2.3. Assumptions with respect to relative air mass calculations and an equation for the relative air mass are discussed in section 2.4.1 Assumptions with respect to the relatively wide spectral response of the LEDs are discussed in section 2.4.2. The conclusions are presented in section 2.5. 2.1 Instrument description In the GLOBE Aerosol Monitoring Project, direct Solar measurements are done with the LED-based Sun photometer developed for the GLOBE project, as first described by Mims [1992]. The instrument is schematically pictured in Figure 2.1. Two LEDs are used as light detectors, one at 58 nm (green), and one at 625 nm (red). A 9 Volt battery is present to supply power to the LEDs. When a beam of light shines on a LED, a current proportional to the light intensity is produced [Mims, 1992]. The current is sent through a resistance and the voltage over the resistance is measured with a voltmeter. The LEDs have a response bandwith of approximately 75 nm (green) and 4 nm (red). The 58 and 625 nm values are effective wavelengths. The concept of effective wavelengths will be discussed in section 2.4.2. The use of a LED as a detector is what distinguishes the GLOBE Sun photometer from professional Sun Figure 2.1. LED-based Sun photometer for the GLOBE project. The detector s opening angle is exaggerated for illustrative reasons. The rectangles with the green and red spheres represent the LEDs. photometers. Advantages of using LEDs are that they are widely available, inexpensive and have stable optical properties. A disadvantage is that their spectral response bandwith is large (up to 75 nm). The implication of this is discussed in section 2.4.2. Professional instruments, on the other hand, use detectors with a spectral response of about 15 nm and they are available in a wide range of wavelengths, but they are expensive and delicate. The cheap GLOBE Sun photometer is therefore very useful for the GLOBE Aerosol Monitoring Project, in which the instrument is used in large numbers. In addition, the robustness of the GLOBE Sun photometer is an advantage since it is used by undergraduate students who cannot be expected to handle the instrument with the same care as experienced scientists. There is an essential difference between the measuring methods of measurements done with professional Sun photometers and measurements done with the GLOBE Sun photometer. A professional Sun photometer is usually placed on an automatic Suntracker which makes it possible to automatically measure AOT with a small, constant sample time during the whole day. A GLOBE Sun photometer measurement requires the physical presence of students, who have to align the instrument every measurement, which introduces a measuring uncertainty because of small alignment errors. On the other hand, their physical presence allows the registering of metadata like aircraft contrails in the light path, cloud cover and haze at the time of measurement, thus providing valuable additional information that professional instrument sometimes lack. 9

2.2 Physical principles The amount of light extinguished by aerosols is determined by both the total aerosol column between the detector and the Sun, and the aerosol absorption and scattering characteristics. In principle, no a priori knowledge about the aerosol type prevailing over the measurement location is available. Since the absorption and scattering characteristics depend on the aerosol type, assumptions on aerosol type have to be made in order to retrieve the total aerosol column. To avoid such assumptions, the total aerosol load in the atmosphere is usually represented by AOT. This dimensionless number represents the light extinction caused by scattering and absorption by aerosols along the light path. The advantage of representing aerosol by AOT instead of the aerosol column is that it is possible to directly compare measurements done at different times and locations without assumptions on aerosol type. When AOT is measured at two wavelengths or more it is possible to retrieve information on aerosol particle size. Scattering and absorption by aerosols usually shows a wavelength dependence according to the empirical relation found by Ångström [1929], which states that AOT (τ a ) decreases with wavelength (λ) as follows, α λ τ ( λ) τ ( λ ) λ a = a, (2.1) where is the so called Ångström coefficient that ranges between and 2.5. For example, the mean value of is 1.398 at De Bilt [Stammes and Henzing, 2]. The exact value of depends on the aerosol particle size distribution. Large aerosol particles (< 1µm), like sea salt and soil dust, generally give rise to a small value for, while small aerosol particles (> 1µm), like soot and sulfates, generally give rise to large values for. From the AOT of two different wavelengths, that are not too close, can be determined and information on particle size distribution can be retrieved. The measuring method of a Sun photometer is based on the Beer-Bouguer-Lambert (BBL) law. The BBL law states that the intensity of a monochromatic beam of Sunlight, I(λ), at the detector is, I τ ( λ ) M ( t ) ( λ) I ( λ) e =, (2.2) with I (λ) the Solar irradiance at wavelength λ at the top of the atmosphere (TOA), (λ) the optical thickness of the atmosphere at wavelength λ and M(t) the relative air mass, that can be interpreted as the light path through the atmosphere and is 1 for overhead Sun (at the Earth s surface at sea level). In Eq. 2.2 it is assumed that the intensity of scattered light at the detector is negligible which is a realistic assumption considering that if the instrument is aligned right next to the Sun the response is zero. If we assume that the instrument gives a response voltage that is proportional to the incoming light intensity at λ, we can replace I(λ) and I (λ) in Eq. 2.2 by V(λ) and V,λ respectively. V, then represents the voltage the instrument would measure at the top of the atmosphere. The value of V, depends on the instrument s characteristics and is called the extraterrestrial constant of the instrument. Since it is an instrument constant the -dependence is dropped, but in general, V, is not equal for the two channels (58 nm and 625 nm) and that is why appears as an index. V, is multiplied by (r /r(t)) 2, where r(t) is the Earth-Sun distance and r is one Astronomical Unit (AU), in order to account for the variations in TOA radiance due to the seasonal variation in the Earth-Sun distance. We write r /r(t) explicitly to illustrate that r is a function of time. The equation relating the instrument s voltage to the optical thickness of the atmosphere then becomes, V r ( λ) r( t ) 2 τ ( λ ) M ( t ) ( ) V e =, λ. (2.3) In principle, V, also varies in time because of fluctuations of I, due to variations in number of sunspots but this effect is smaller than.2 % [Liou, 22] and therefore neglected. 1

The total optical thickness of the atmosphere, (λ), is the sum of Rayleigh scattering optical thickness, R (λ), optical thickness due to absorption by gases, g (λ), and optical thickness due to scattering and absorption by aerosols, a (λ). If (λ) is known and Rayleigh scattering and absorption by gases can be corrected for, a (λ) can be retrieved. In order to calculate (λ) and consequently a (λ) from Eq. 2.3, knowledge of V(λ), V,λ and M(t) is required. V(λ) is the actual measurement. The determination of V,λ is discussed in the next chapter and the calculation of M(t) from the Solar elevation angle is discussed in section 2.4.1. Apart from Rayleigh scattering, absorption by gases and absorption and scattering by aerosols, there is also scattering and absorption by clouds when present. In practice it is impossible to distinguish the light extinction due to aerosols from the light extinction by clouds. Therefore, AOT retrieval is possible only when the Sun is not obscured by clouds. Recognising cloud-free conditions and correctly reporting cloud conditions is therefore essential for reliable AOT retrieval. 2.3 Measurement method In order to retrieve a (λ) from Eq. 2.3 we need to know V,λ. This can be done with the Langley method [Liou, 22]. The idea of the Langley method is to measure V() for a wide range of relative air masses (or Solar zenith angles). A plot of ln(v()) versus M(t) may be extrapolated to the zero point, which represents the top of the atmosphere (M(t) = ). This method has the advantage of being easy, relative to measuring I, at the top of the atmosphere. Taking the logarithm on both sides of Eq. 2.3 it is clear that, if the BBL law can be applied, the Langley plot should be a straight line, r ( V ( )) ln r( t ) 2 ( ) V ) ( λ) M ( t) ln λ τ λ. (2.4) =, The slope of the Langley plot is τ(λ), the total (time-averaged) optical thickness of the atmosphere and the zero point (M(t) = ) is ln((r /r)v,λ ). For an accurate determination of V, the optical thickness of the atmosphere should be constant during the day of measurement. This condition can be satisfied by taking the Langley plot measurements on a clear day with constant (and therefore preferably low) AOT. Furthermore the Langley plot method is only valid if the BBL law holds, that is if the instrument detects a monochromatic beam of light. Since all instruments detect over a finite bandwith this is an approximation that should be investigated. The validity of the Langley plot method for the 75 nm broadband LED-based Sun photometer is discussed in section 2.4.2. If a particular Sun photometer s extraterrestrial constant V,λ is known, other instruments can be calibrated by determining response ratios to the calibrated instrument, which is used as a reference instrument. Response ratios are determined by comparing simultaneous measurements of the reference instrument and the instrument under calibration. This method saves the time-consuming Langley plot measurements for every individual instrument and is very useful for the GLOBE Aerosol Monitoring Project, in which 2 GLOBE Sun photometers are in use. When V,λ is known, we obtain the equation for the optical thickness of the atmosphere (λ) by rearranging terms in (Eq. 2.4) τ ( λ) 1 = M ( t) ln r ( ) r ( t ) V 2 V, λ ( ) λ. (2.5) In order to determine the contribution a (λ) to the total optical thickness, (λ), we subtract Rayleigh optical thickness R (λ) and gaseous absorption optical thickness g (λ). Within the spectral domain of the GLOBE Sun photometer the only significant gaseous absorber is ozone, so g (λ) is approximated by o3 (λ). The equation for the retrieval of AOT now becomes, 11

τ a ( λ) 1 M r ( t ), λ d = ln τ λ τ λ. (2.6) ( t) r 2 ( ) ( V V ) V ( λ) V d R ( ) ( ) o3 2.4 Assumptions In this section we investigate two assumptions that are important in calculating AOT from GLOBE LED-based Sun photometer measurements. In Eq. 2.6 it is assumed that M(t) is the same for a, R and o3. The value of M(t) depends on the vertical distribution and since, in principle, aerosols, molecules and ozone have different vertical distributions, M(t) should be calculated separately for each component. The assumption of taking one value for M(t) is discussed in section 2.4.1. In Eq. 2.6 the monochromatic BBL law is applied to LEDs, that have wide spectral responses. The assumption that the monochromatic BBL law is valid for the wideband LEDs is discussed in section 2.4.2. 2.4.1 Relative air mass and vertical distribution The relative air mass is usually interpreted as the number that quantifies the average photon path from the Sun through the atmosphere to the detector. In fact it is a number that represents the amount of scattering and absorption along the light path, relative to that for the Sun at zenith. Figure 2.2 shows the relative air mass, M(t) at zenith angle θ = θ and at zenith angle θ =. The light beam at zenith angle = θ in Figure 2.2 shows that the average photon path through stratospheric species, like ozone, is smaller than the average photon path through tropospheric species, like air molecules and aerosols. The light beam at = shows that this effect does not appear at =. From this we can conclude that for large zenith angles the relative air mass in principle is different for species with different vertical distributions, while for small relative air mass this effect is negligible. Eq. 2.6 assumes that for the different components of the optical thickness the relative air masses are the same, that is, e M R ( t ) τ R ( λ ) M o3 ( t ) τ o3 ( λ ) M a ( t ) τ a ( λ ) M ( t ) [ τ R ( λ ) + τ o ( λ ) + τ a ( λ )]. (2.8) e e 3 = e Results from Thomason et al [1983] show that errors arising from different relative air masses from different vertical distributions are significant only at very large Solar zenith angles ( > 7 ). Since almost all measurements are done at Solar zenith angles smaller than 7º the errors in AOT arising from wrong values for M o3 (t) are neglected, noting that it introduces a maximum error of.1 AOT at large AOT and at large Solar zenith angle ( > 7 ). Note that, since aerosol is assumed to peak in the troposphere, these results are applicable to measurements of tropospheric aerosol only. When large amounts of aerosol enter the stratosphere by vulcanic eruptions, errors in relative air mass may effect AOT results significantly. 12

Figure 2.2. Schematic representation of the effect of vertical distribution on relative air mass. The grey bars represent air molecules and aerosols, whose concentrations peak in the troposphere. The green bars represent ozone, whose concentration peaks in the stratosphere. In practice, ozone, air molecule and aerosol concentrations have a more smooth profile. An effect that also influences the relative air mass is refraction of the light beam in the atmosphere. The effect of refraction on relative air mass is shown in Figure 2.3. In order to determine an upper limit for errors arising from neglecting internal refraction into account, the difference in relative air mass calculations is determined at large Solar zenith angle, θ TOA = 7. and with the atmosphere regarded as one single layer witn n sealevel = 1.28. Using Snellius law of refraction, n 1 sin(θ 1 ) = n 2 sin(θ 2 ), this yiels θ app = 69.956. This means that the errors in M(t) that are introduced when internal refraction is not taken into account are smaller than.1 and therefore this effect is not taken into account. Figure 2.3. Internal refraction of a lightbeam in the atmosphere. 13

Now we can introduce an equation for M(t) without taking into account vertical distributions. The relative air mass is calculated from the Solar zenith angle which is a function of time, that is = (t). The relative air mass is often approximated by sec(θ) which is called the geometric air mass, and in which the curvature of the Earth s surface and internal refraction in the atmosphere are not included. An equation for the relative air mass that takes into account the curvature of the Earth s surface and internal refraction in the atmosphere is taken from Young [1994], 2 1.2432cos θ +.148386cosθ +.96467 M ( θ ) =. (2.9) 3 2 cos θ +.149864cos θ +.12963cosθ +.33978 The difference between M(t) calculated using Eq. 2.9 and sec(θ) is shown in Figure 2.4. This shows that for small zenith angles sec(θ) is a very good approximation, but for θ > 7 the difference between sec(θ) an Eq. 2.9 is significant (>.5 %). This means that for Solar zenith angles larger then 7 the effect of a spherical Earth has a significant effect on the relative air mass and must be taken into account. 2.4.2 Finite bandwith and BBL It is all but trivial that the monochromatic BBL law can be applied to the broadband LEDs. When a detector of finite bandwith is treated as a monochromatic detector, this introduces errors that should be investigated. In professional Sun photometers, light detectors have a bandwith of typically 15 nm or less. The difference in AOT, calculated from a detector with a 15 nm bandwith with central wavelength eff and from a monochromatic detector at eff, is typically.5 AOT. Therefore, professional Sun photometer detectors are often treated as monochromatic. The LEDs in the GLOBE Sun photometer have a bandwith of 75 nm (58 nm) and 4 nm (625 nm). A LED measures light over a wide spectral range. In order to get AOT results at a certain wavelength, an effective wavelength for a LED is defined. Following Brooks and Mims [21], the effective wavelengths for a LED with spectral response function R(λ) is defined as, R( λ) I λeff =, (2.1) R( λ) I ( λ) dλ Figure 2.4. Relative air mass according to Young s formula and according to the geometrical opproximation (sec). ( λ) λdλ where I (λ) is the extraterrestrial Solar irradiance at wavelength λ. The Solar irradiance at TOA, I, is taken rather than the radiance at the detector opening, I, in order to obtain a λ eff that is independent of the conditions at the measuring site. Since the wavelength dependence of the measured radiance at the detector depends on AOT, Ångström coefficient, pressure, Ozone column and elevation of the measuring site, defining the effective wavelength at the detector (replacing I in Eq. 2.1 by I) would result in a unique effective wavelength for every measurement. We note that by defining the effective wavelength we may introduce errors, whose extend is to be verified in future work. 14

The LED s output voltage depends on the amount of Rayleigh scattering, ozone absorption and Mie scattering by aerosols. The wavelength dependences of Rayleigh and Mie scattering cross sections, ozone absorption cross section and the response functions of the LEDs used in the GLOBE Sun photometer (HLMP-D6 and HLMP-3762) are plotted in Figure 2.5. The Rayleigh scattering cross section (solid curve) shows a λ -4 dependence. The Mie scattering cross section (dashed line) shows a wavelength dependence, dependent on the aerosol particle size, in the λ -1 - λ -2 range. The ozone absorption cross section (dotted curve) shows a capricious behavior. Figure 2.5 shows that the BBL law cannot simply be applied since Rayleigh, Mie and ozone cross sections have different spectrally weighted contributions at the detector. Note that the height of the curves is arbitrary and does not contain any information on the amount of scattering and absorption. Figure 2.5. Wavelength dependence of Rayleigh scattering ( -4 ) (solid curve), Mie scattering and absorption ( -1 and -2 ) (stripes), ozone absorption (dotted curve) and the response functions of the HLMP-D6 LED (left) and the HLMP-3762 LED (right). To account for the LED s large bandwith, the monochromatic optical thickness values that are calculated from the cross section (σ) and total column (N), i ( λ) = σ i ( λ) N i τ, (2.11a) should be replaced by effective values. Effective optical thickness values R,eff and o3,eff for the LED s with response function R(λ) can be defined by, R( λ) I ( λ) λ τ i, eff =. (2.11b) R( λ) I ( λ) dλ λ ( λ) τ In general the value of i,eff is not equal to the monochromatic value of i at λ eff. The effective values for Rayleigh scattering and ozone absorption at standard atmospheric conditions, that are calculated using Eq. 2.11b for the HLMP-D6 and HLMP-3762 LEDs, are listed in Table 2.1. The monochromatic Rayleigh scattering coefficients and ozone absorption coefficients for the two LEDs at the LED s effective wavelengths, that are calculated from Eq. 2.11a, are also listed in Table 2.1. Table 2.1 gives an idea of the impact of the LED s large bandwith on the Rayleigh and ozone contributions to the total optical thickness. i dλ 15

LED τ R,eff τ R at λ eff τ o3,eff τ o3 at λ eff (eq.2.11b) (eq.2.11a) (eq.2.11b) (eq.2.11a) HLMP-D6 (λ eff = 58 nm).145.135.13.13 HLMP-3762 (λ eff = 625 nm).6.58.29.31 Table 2.1. Rayleigh scattering and ozone absorption optical thickness at standard atmospheric conditions, effective values and monochromatic values at eff. The effective values for Rayleigh scattering using Eq. 2.11b and the effective values for ozone absorption using Eq. 2.11b show significant differences with the values one would obtain by taking the monochromatic optical thickness at λ eff, using Eq. 2.11a. Using monochromatic optical thickness at λ eff would result in systematic errors of about.6 AOT, and therefore the effective values, using Eq. 2.11b, are used to calculate AOT. The transmittance of light through the atmosphere, detected by the GLOBE Sun photometer, should not be represented by the monochromatic attenuation e -()M(t) (Eq. 2.2), but should be represented by a transmittance function T, defined by, V = V * T, (2.12) which accounts for the full spectral range averaged by the detector. The transmittance function T is equal to the normalized spectral response of the detector, T true = R( λ) I ( λ) e M R( λ) I ( t ) [ τ ( λ ) + τ ( λ ) + τ ( λ ) R a ( λ) dλ 3 dλ ]. (2.13a) Alternatively, a transmittance function using effective values can be defined, based on Eq. 2.2 and Eq. 11b, M ( t ) [ τ R, eff + τ o3, eff + τ a, λ ] eff =. (2.13b) T eff e If these two transmittance functions give comparable results for V in Eq. 2.12, that is T eff T true, then Eq. 2.13b, is a good approximation for 2.13a and τ a,λeff is a good measure for the aerosol optical thickness at λ eff. In order to investigate the differences in AOT results when T true is approximated by T eff, the behavior of the LEDs is simulated in Brooks and Mims [21]. The simulation is done by calculating a Langley plot based on Eq. 2.13a with assumed aerosol optical thickness of.1 AOT at 58 nm. The Ångström coefficient that was used is not denoted in the article. A second Langley plot is calculated based on Eq. 2.13b (effective values). The two Langley plots are shown in Figure 2.6. 16

ln(v), simulated, -,5 -,1 -,15 -,2 -,25 -,3 -,35 -,4 -,45 -,5 1 2 3 4 5 6 7 Relative air mass Figure 2.6. Simulated Langley plot calibrations by Brooks and Mims [21] at.1 AOT at 58 nm. Solid line with markers: based on Eq. 2.13a. Dotted line: based on Eq. 2.13b. At relative air masses smaller than 6 (which comes down to Solar zenith angle of 8 ) the differences are negligible (< 1%), meaning that the effective value-approach can be used in that regime. This implies that the transmittance function for M(t) < 6 can be written as Eq. 2.13b, that is that the monochromatic BBL law can be applied to the broadband LED for relative air masses smaller then 6 and at.1 AOT. Since the wavelength dependence of the measured radiance at the detector depends on both AOT and Ångström coefficient, approximating Eq. 2.13a with Eq. 2.13b should also be done for a range of higher AOT values and at several values of α. This is something that should be done in future work. 2.5 Conclusion The main conclusion of this chapter is that the monochromatic BBL law can be applied to the broadband LED-based Sun photometer. Effective values for Rayleigh scattering optical thickness and ozone absorption optical thickness should be used in AOT calculations to avoid systematic errors arising from the LED s wide spectral response. Using the geometrical air mass approximation will lead to systematic errors. Instead, the relative air mass should be calculated using Young s formula. The effect of different vertical distributions on relative air mass is negligible and therefore one single relative air mass is used for in AOT calculations. AOT measurements with the LED-based GLOBE Sun photometer are meaningful when measurements are done at relative air masses less then 6. 17

3 Instrument Calibration For AOT calculation from GLOBE Sun photometer measurements the instrument s extraterrestrial constants need to be known. Since the aerosol retrieval method is sensitive to errors in the extraterrestrial constants (see chapter 4) this calibration of the instrument should be done with the highest accuracy. In order to calibrate the GLOBE Sun photometer used at KNMI (serial number: RG2-47), a Langley plot analysis is done. After calibration, RG2-47 is used as a reference instrument for relative calibration of several other instruments. The Langley plot analysis of RG2-47 is discussed in section 3.1 and the results of the relative calibrations are summarized in section 3.2. For a particular Sun photometer an accurate Langley plot calibration or a relative calibration to obtain the instrument s extraterrestrial constant V,λ is needed only once when degradation of the LED and fluctuations of I () due to variations in number of Sunspots are not considered. 3.1 Langley analysis of RG2-47 A Langley plot is a plot of the instrument voltage versus relative air mass (M(t)). The extraterrestrial constant can be found by extrapolating a linear least squares fit to M(t) =, that is the top of the atmosphere. A Langley plot requires constant optical thickness during the measurement period and therefore stable atmospheric conditions are needed. Low AOT values are desirable, since aerosol concentrations are expected to be little variable at low AOT. Furthermore a Langley plot requires a sufficient large relative air mass range. On several days, Langley plot measurements are done for RG2-47. The measurements performed in the morning of April 8, 23 are used for the final calibration. This day was the best for a Langley plot analysis for the following reasons: 23-4-8 was a day with relatively low AOT (.1 at 58 nm). AOT was stable during the measurement time (±.3). The measurements are on a straight line in the Langley plot. The calibration constants from these measurements give realistic AOT results (AOT 58nm > AOT 625nm ). Results from the Brewer spectrofotometer at KNMI [Allaart et al, 2] show a stable atmosphere with good Langley-plot conditions on 23-4-8 (Figure 3.2). The Langley plots from the 23-4-8 measurements are shown in Figure 3.1. Figure 3.1. Langley plot measurements for RG2-47 taken at 23-4-8. Figure 3.2. Brewer measurements (363 nm) at 23-4-8 (Figure by M.Allaart, KNMI). The extraterrestrial constants resulting from the linear extrapolation are 2.323 V ±.5 (58 nm) and 1.899 V ±.5 (625 nm). Figure 3.2 shows the Brewer 363 nm Langley plot at 23-4-8. The points indicated by the arrows are measurements done in the morning, when the Langley plot measurements for RG2-47 are also done, and they lie on a relatively straight line. The other 18

measurements are taken in the afternoon and they show no structure because of the presence of clouds. In the morning ln(v) shows a linear dependence on M(t), i.e the optical thickness is constant (at 363 nm). The uncertainty in the GLOBE V,λ s caused by the uncertainty in the linear fit is 6 mv. However, V is estimated 5 mv based on results for V, from other Langley plots. This uncertainty is regarded as an upper limit for the uncertainty in V,λ. The 5 mv estimation is obtained by comparing results from all Langley plots that are made for RG2-47. 3.2 Relative calibrations All the other Sun photometers that are in use in the GLOBE Aerosol Monitoring Project in the Netherlands are calibrated relative to RG2-47. Simultaneous measurements of the instrument to be calibrated and RG2-47 are compared in order to find the response ratio, R. The simultaneous measurements are done at a range of atmospheric conditions in order to cover an as large as possible voltage range of the instrument. As an example, the resulting relative plot for RGK-24 is shown in Figure 3.3. From a linear fit to the data, R is obtained. The extraterrestrial constants are now calculated using V, λ ( instr.) = V, λ ( RG2 47) R. (3.1) The calibration constants for the Sun photometers obtained with Eq. 3.1 are listed in Table 3.1. The precision of the V,λ s is estimated to be 5 mv. This number comes from the uncertainty of the V,λ of the reference instrument. The uncertainty in the fitting parameters is much smaller than 5 mv. Serial number C instrument (V) 625 C instrument (V) 58 user RG2-47 1.899 2.323 KNMI RGK-21 1.579 1.96 BN-college RGK-22 1.443 1.44 Mozaiek RGK-23 --------- --------- ------------------- RGK-24 1.44 1.867 Alkwin College Figure 3.3. Relative calibration of RGK-24. RGK-25 --------- --------- ------------------- RGK-26 1.385 2.361 Populier RGK-27 1.23 1.681 ------------------- RGK-28.627 1.385 Ichthus College RGK-29 1.332 1.848 Pascal college RGK-21.956 2.48 SG. Tabor (Locatie Oscar Romero) RGK-211 1.312 2.391 ------------------- RGK-212 1.289 2.27 Goois Lyceum RGK-214 1.36 2.3 Zwin College Table 3.1. Calibration constants and users of GLOBE Sun photometers used in the GLOBE Aerosol Monitoring project. 3.3 Conclusion The GLOBE Sun photometer at KNMI (RG2-47) was calibrated by analysis of Langley plot measurements performed on 23-4-8. All instruments used in the Dutch GLOBE Aerosol Monitoring Project have been calibrated relative to RG2-47. This, together with the fact that AOT results are sensitive to errors in the extraterrestrial constants, is a reason to proceed taking calibration measurements for RG2-47 to improve the precision and accuracy of the calibration. 19

4 Algorithm The GLOBE program provides for an on-line algorithm that allows quick and simple processing of GLOBE Sun photometer data. In this work, an algorithm is developed independently from the GLOBE algorithm, which makes processing of data from GLOBE Sun photometer measurements possible at KNMI. The development of an algorithm at KNMI is essential for the GLOBE Aerosol Monitoring Project for the following reasons: 1. An independent algorithm makes it possible to compare the algorithm with the GLOBE algorithm, which may lead to improvements. 2. The development of an algorithm improves the error analysis, and thus improves the quantitative estimate of the uncertainty associated with AOT results at KNMI. 3. The development of an algorithm gives insight into the sensitivities of AOT calculation. The equation governing the AOT calculation from the instrument s voltage is presented in section 4.1 and is tested by comparing results with results from the on-line GLOBE algorithm. The KNMI GLOBE algorithm comparison is presented in section 4.2. The error analysis is discussed in section 4.3 and the conclusions are presented in section 4.4. 4.1 AOT Calculation Following Eq. 2.6 with the effective values for the Rayleigh and ozone coefficients from Table 2.1 and subtracting the instrument s dark voltage V d (that is the voltage obtained from the instrument in total darkness) from V and V,, we obtain the equation used in the algorithm to retrieve AOT from GLOBE Sun photometer measurements, τ a ( λ ) eff 1 M t r( t ), λ d =, (4.1) ( ) ln r 2 ( ) ( V V ) V λ eff V d τ R, eff ( p) τ ( N) where N denotes the overhead ozone column in Dobson Units (DU), p denotes the pressure and we write M(t) explicitly to illustrate the that the relative air mass is a function of measurement time. We will discuss all the elements in the equation: M(t) is the relative air mass which depends on the time of measurement. M(t) is calculated from the Solar zenith angle,, using Eq. 2.9. The Solar zenith angle is calculated using cos() = sin()sin()+cos()cos()cos(h) where is the latitude, the declination of the Sun and h is the hour angle. The declination of the Sun is calculated using the equation given by Liou [22]. Measurements with M(t) > 6 are rejected because systematic errors become significant at M(t) > 6 (see section 2.4.2). (r /r) 2 accounts for the fluctuations in TOA irradiation due to the non-spherical orbit of the Earth around the Sun. Its value depends on the measurement time and has a one-year period in which it ranges between 1.34 on January 3 and.967 on July 5 [Liou, 22]. V, is the extraterrestrial constant of the instrument s channel. It is determined by a Langley plot analysis that is discussed in section 3.1. The extraterrestrial constant is determined for an Earth-Sun distance of 1 AU, so V, is interpreted as the voltage the instrument would measure at TOA at 1 AU. The value of V, depends on the components of the instrument. V,58 is in general not equal to V,625. For most Sun photometers the values of the extraterrestrial constants range between 1 mv and 22 mv. V d is the dark voltage, that is the instrument voltage in total darkness. It is measured by sealing off the detector opening and denoting the output voltage. It is always less than 2 mv. V eff is the instrument voltage, the actual measurement. It is obtained by pointing the Sun photometer opening at the Sun, aligning it, and denoting the maximum output voltage in a period of 1 seconds for a cloud-free Soalr disk. Its value depends on the relative air mass and the total optical thickness of the atmosphere. Its value is somewhere between V d and V, except for values close to V because there is always a certain amount of Rayleigh scattering and ozone absorption present in the lightpath. o3, eff 2

The Rayleigh scattering effective optical thickness, R,eff (p), is calculated from the effective coefficients in Table 2.1 and the surface pressure, τ ( = p p τ. The standard R, eff p ) R, eff pressure, p, is taken 113 mbar. The pressure is obtained by KNMI meteo data for measurements done at KNMI, and by a barometer for GLOBE students. At sea level, the pressure normally ranges between 98 and 14 mbar in the Netherlands, which makes the effective Rayleigh coefficients about.14 (58 nm) and.6 (625 nm). Since AOT generally ranges between.5 and.5 Rayleigh scattering is a relatively large component of the optical thickness at 58 and 625 nm. The ozone absorption effective optical thickness, o3,eff (N), is calculated from the effective coefficients in Table 2.1 and the ozone column, τ ( ) o, eff N = N N τ The standard 3 o 3, eff ozone column, N, is taken 3 DU. The total ozone column is obtained by Brewer UV ozone measurements for measurements done at KNMI [Allaart et al, 2]. For measurements done by GLOBE students the ozone column is determined at KNMI by extrapolating Brewer UV ozone measurements to the GLOBE school measuring location. Since the ozone column over the Netherlands can have significant gradients this extrapolating introduces errors that should be accounted for in the error analysis. However, the ozone column gradient is usually small at clear and stable atmospheric conditions at high atmospheric pressure, when most GLOBE AOT measurements are done. Normally the ozone column ranges between 25-4 DU which makes the effective ozone absorption coefficients range between.1 and.2 (58 nm) and between.2 and.4 (625 nm). Ozone absorption contributes less to the optical thickness than Rayleigh scattering, but the ozone absorption contribution is a significant component and should therefore be included. The metadata is written down directly after the measurement. This is the data that describes the conditions during the measurement and contains information on temperature, cloud cover, cloud types, sky color, haziness and all other reasons for obscured sky. Accurately registering metadata provides additional value to the GLOBE AOT measurements that AOT measurements done with a fully automatic instrument usually lack. The check with metadata can be used to expel measurements performed under bad conditions and is therefore an essential feedback system in the measurement method of the GLOBE Sun photometer. The algorithm that is developed in this work is shown in a flowchart in Figure 4.1. The check with the metadata is essential in the decision to accept or reject data and is represented by the dashed line. If the data is rejected, then it is not stored. Eq. 4.1 assumes that the only significant absorber within these two wavelength ranges is ozone. H 2 O also absorbs in this wavelength regime. The H 2 O absorption cross-section has order of magnitude 11-26 moleculescm -2 for both channels. Over the Netherlands the water vapor column ranges between - 35 kgm -2 [De Haan and Barlag, 23]. This gives a maximum contribution of H 2 O to the optical thickness of 11-3 for measurements done in the Netherlands. This is very small compared to Rayleigh scattering and ozone absorption and therefore it is not included in the algorithm. 21